3 hours ago, aljude said:

I run an Intel i9-7900X and I use Lasso, you can disable HT and select your preferred cores so I have everything bar games on cores 0 to 2, then games on all other cores 3 to 19... apart from P3Dv5 which is set to run on even cores 4 to 18...

It is better to use an AM inside the P3D.cfg file as the P3D then controls the preferred cores from startup.

PL is outside P3D and controls the cores from outside. If you search on PL you will find tests that show that an AM inside P3D performs better.

The same goes for Addons : using a batchfile to control the addon cores is working better as the Addons are started on the preferred cores instead of being forced onto cores.

I have used PL too in the past.

Just now, GSalden said:

It is better to use an AM inside the P3D.cfg file as the P3D then controls the preferred cores from startup. PL is outside P3D and controls the cores from outside.

I'd bet you a dollar that they do it in the exact same Windows API call, and if it's not the same, the effects are - the kernel scheduler copies that bitmap to an internal structure and uses it to dispatch threads.

Has anyone ever determined exactly what routine in P3D is blocking when a stutter happens?

Cheers

I have an I9 9900k and tried many times

with HT on with affinity mask 21845 and no

affinity mask att all and get more stutters 

than with HT off and much hugger temps also.

 

On 4/19/2020 at 8:34 PM, Ianrivaldosmith said:

I think I finally just figured mine out now. 
I run affinity mask so that only 8 cores are on (disabling HT, via the mask). 

Why do you keep saying you disable HT? No, you don't disable HT, technically you just use a specifc AM that excludes second LPs on all cores for only one app running (among dozens of others enjoying HT) Your system overall might still benefit from HT On. I hope you're not an HT hater, are you? HT doesn't sting and generally is good for your health, lol 😄(sorry, couldn't help) But you know that already since you don't really turn it off. Same here btw with a similar AM now.

Cheers.

Edited August 16, 20205 yr by Dirk98

16 hours ago, Luke said:

I'd bet you a dollar that they do it in the exact same Windows API call, and if it's not the same, the effects are - the kernel scheduler copies that bitmap to an internal structure and uses it to dispatch threads.

Has anyone ever determined exactly what routine in P3D is blocking when a stutter happens?

Cheers

Search on the Avsim forums : Steve W has shown how and several times.

17 hours ago, Dirk98 said:

I hope you're not an HT hater, are you? HT doesn't sting and generally is good for your health, lol 😄

Iam not a hater but have run HToff very often in P3D, 

for now i have P3D v4.5 and V5 on my 9900k run AM 65365 , 65493 , 65525 all work great 62805 work to 

today 65525

Edited August 17, 20205 yr by westman

6 hours ago, GSalden said:

Search on the Avsim forums : Steve W has shown how and several times.

What search term should I use?

V5 is another beast for i7-4770K than P3d prior versions and performance is much better with HT ON.

First time i am using it now with V5

I think its the same to all 4-core cpus

Michael Moe

Edited August 17, 20205 yr by Michael Moe

2 hours ago, Luke said:

What search term should I use?

I searched "stutter" by author (SteveW) and got 107 results.

Hope that helps.

Jesse

2 hours ago, Michael Moe said:

V5 is another beast for i7-4770K than P3d prior versions and performance is much better with HT ON.

First time i am using it now with V5

I think its the same to all 4-core cpus

Michael Moe

You are right,

Both my 7940X ( 14 cores ) and 8700K ( 6 cores ) are running V5 with HT off.

1 hour ago, GSalden said:

You are right,

Both my 7940X ( 14 cores ) and 8700K ( 6 cores ) are running V5 with HT off.

Lol, If he's right then you are running V5 with HT On in context. But whatever makes you happy, Gerard, of course.

Cheers.

 

Edited August 17, 20205 yr by Dirk98

18 minutes ago, Dirk98 said:

Lol, If he's right then you are running V5 with HT On in context. But whatever makes you happy, Gerard, of course.

Cheers.

 

Hi Dirk, I mean that he is right that HT On on a 4 core CPU is better for v5.....

However, I have more cores 😄

 

One idea for stutter is that when the simulator flight arrives somewhere pulling in more data, the intensity of tasks saturates the CPU. Let's say we have ten cores and HT off no AM. All the cores are going to be grabbing as much throughput as they can, which impacts the system overall, and can pull down the main rendering task. We can use an AM to disallow the last one or two cores, e.g. 0011111111=255. That way those cores unused by the simulator background task are free to attract system throughput. If we go to HT enabled and as Dirk says enable only one LP of each core, e.g. 01,01,01,01,01,01,01,01,01,01 we still use ten cores and when the sim pulls in  new data we can see activity build on LP1, LP0 having the first sim task is now sharing a core with that load on LP1. Go to AM 00,00,01,01,01,01,01,01,01,01 with only 8 sim cores and have a look at the Task Manager graphs and see how LP1 attracts less activity. That would enable the simulator rendering to remain more consistent. With less cores available we can still disable one or two cores, and add more throughput with the HT mode as Gerard says, by doubling LPs used on the later cores.

Always use LP0 (or core zero HT Off) in the AM since this allows the jobscheduler to keep the sim tasks up to date. With HT enabled and an AM that ignores LP0 we can see in a fast flight and flying to a new area very quickly, the Autogen takes longer to catch up. With LP0 enabled in the AM we get to the new area and the Autogen is rendering as normal. I'm not certain that happens an all systems but it can be a problem of autogen missing..

3 hours ago, SteveW said:

One idea for stutter is that when the simulator flight arrives somewhere pulling in more data, the intensity of tasks saturates the CPU. Let's say we have ten cores and HT off no AM. All the cores are going to be grabbing as much throughput as they can, which impacts the system overall, and can pull down the main rendering task. We can use an AM to disallow the last one or two cores, e.g. 0011111111=255. That way those cores unused by the simulator background task are free to attract system throughput. If we go to HT enabled and as Dirk says enable only one LP of each core, e.g. 01,01,01,01,01,01,01,01,01,01 we still use ten cores and when the sim pulls in  new data we can see activity build on LP1, LP0 having the first sim task is now sharing a core with that load on LP1. Go to AM 00,00,01,01,01,01,01,01,01,01 with only 8 sim cores and have a look at the Task Manager graphs and see how LP1 attracts less activity. That would enable the simulator rendering to remain more consistent. With less cores available we can still disable one or two cores, and add more throughput with the HT mode as Gerard says, by doubling LPs used on the later cores.

Always use LP0 (or core zero HT Off) in the AM since this allows the jobscheduler to keep the sim tasks up to date. With HT enabled and an AM that ignores LP0 we can see in a fast flight and flying to a new area very quickly, the Autogen takes longer to catch up. With LP0 enabled in the AM we get to the new area and the Autogen is rendering as normal. I'm not certain that happens an all systems but it can be a problem of autogen missing..

Couldn’t I just have HT on no AM

go into process lasso got to each of my addons uncheck all off the first 2 cores ??

and done?

thanks

mike

No.

You could uncheck the two LPs of the first two cores but you should use the AM if an app gives you the possibility.